A Cell Counting Method For Microscopic Images Of Biological Tissues

Cell counting is critically important for functional studies and clinical pathology ofbiological samples. At present, flow cytometry is widely used for cell counting; however,the spatial information of the cells can hardly be got. Another method based on imageprocessing has been increasingly popular. Images of biological samples are firstly takenunder microscopes; afterwards, proper image processing techniques are applied for cellcounting. Accordingly, both the cell number and the cellular spatial information areobtained.In the fluorescence images, isolated and adhesive cells are both included. The formeris easy to count, and the key for cell counting is to count adhesive cells. The features ofimages processed in this thesis are listed as following: three cells at most adheringtogether, low proportion of adhesive cells, and large dataset of samples.In view of the features of the dataset, an algorithm based on the concavity point wasemployed in this thesis. Counting of adhesive cells was achieved according to therelationship between the concave point and cell number, with high accuracy ratio andoperating speed. Main procedures of this algorithm include preprocessing, cell counting,data output and estimation. For preprocessing, the images were filtered to decrease thenoises, and the image contrast was increased. For estimation of cell counting, the accuracyratio, omission ratio and loss ratio were calculated.Compared to manual counting, the accuracy ratio, omission ratio and loss ratio of ouralgorithm turned out to be90.6%,3.4%and9.4%, respectively.